the geomorphic view of networking an abstract model of
play

THE GEOMORPHIC VIEW OF NETWORKING: AN ABSTRACT MODEL OF NETWORK - PowerPoint PPT Presentation

THE GEOMORPHIC VIEW OF NETWORKING: AN ABSTRACT MODEL OF NETWORK ARCHITECTURE Jennifer Rexford Pamela Zave Princeton University AT&T LaboratoriesResearch Princeton, New Jersey, USA Florham Park, New Jersey, USA THE CLASSIC


  1. THE GEOMORPHIC VIEW OF NETWORKING: AN ABSTRACT MODEL OF NETWORK ARCHITECTURE Jennifer Rexford Pamela Zave Princeton University AT&T Laboratories—Research Princeton, New Jersey, USA Florham Park, New Jersey, USA

  2. THE “CLASSIC” INTERNET ARCHITECTURE this architecture has succeeded (beyond most peoples’ wildest dreams) in fostering APPLICATION LAYER innovation and shaping the world we live in TRANSPORT LAYER the trend is toward however, it is now a more pluralistic widely agreed that architecture . . . it does not meet NETWORK LAYER society’s present and . . . with multiple, future requirements customized protocol stacks security LINK LAYER dependability mobility PHYSICAL LAYER scalability quality of service resource management

  3. A REAL EXAMPLE headers in a typical AT&T packet (12 instead of 4) Application HTTP HTTP being used as a transport protocol because it is the only TCP way to traverse NAT boxes and firewalls IP IPsec security IP cellular service GTP (mobility, QoS, billing) UDP 15 + load-balancing algorithms operate on this packet, most of IP them understood and tested only in isolation! MPLS multiple layers of MPLS resource management Ethernet

  4. WHAT IS NEEDED FOR THE PLURALISTIC INTERNET? COMMUNICATION SERVICES we need . . . Application a broader range of services HTTP security and resource policies TCP appropriate to each application IP . . . so that all applications can be developed easily and run efficiently IPsec IP SOFTWARE DEVELOPMENT GTP we need to develop all this custom software rapidly and correctly, UDP through code generation and re-use IP NETWORK MANAGEMENT MPLS we need design principles that MPLS enable us to manage complexity, ensure robustness, and predict Ethernet global behavior

  5. A NECESSARY FIRST STEP AN ABSTRACT MODEL OF NETWORK ARCHITECTURE . . . . . . would enable us to describe networks, protocols, and other solution mechanisms . . . . . . in a way that is . . . simple modular unique formalizable comprehensive . . . so that all the ideas and artifacts of networking can be compared and composed

  6. OUTLINE 1 THE “GEOMORPHIC VIEW”: “ AN ABSTRACT MODEL OF NETWORK ARCHITECTURE 2 THE GEOMORPHIC VIEW OF MOBILITY a Understanding mobility b Comparing existing mobility protocols c Generating and composing new mobility mechanisms 3 BEYOND MOBILITY: FUTURE WORK

  7. CLASSIC LAYERS OR THE GEOMORPHIC VIEW OSI REFERENCE MODEL OF NETWORKING there is a fixed number of levels there can be any number of levels some layers have small or local the scope of each layer is global, scopes so layer = level each layer is a microcosm of each layer/level has a specialized networking, containing all the basic function components (state components and mechanisms)

  8. A NEW LAYER MODEL: MEMBERS, ROUTING, AND FORWARDING a process, which is merely a locus of member state and control unique and with the potential for name persistent autonomous action within the layer B D E a LAYER communication link A C channel enables members to send messages to one another, forwarding protocol using the links routes often there is not a link between every pair of members; routes tell the forwarding protocol how to reach one member from another over the existing links, with forwarding by intermediate members routing algorithm maintains the routes as links change over time

  9. A NEW LAYER MODEL: COMMUNICATION SERVICES channel an instance of a communication service session a communication channel (as are links) session protocol implements an end-to-end communication service, on top of the basic, fundamentally unreliable, message delivery provided by the forwarding protocol session LAYER e b a d link from the perspective of one layer, sessions are more convenient than links they have longer reach; might be more reliable, better-behaved (with FIFO delivery), with guaranteed performance, etc.

  10. A NEW LAYER MODEL: THE “USES” HIERARCHY when an overlay uses an underlay, registration a link in the overlay is implemented by a session in the underlay relates an overlay member to the underlay member that it is using on OVERLAY the same machine E (higher A link layer) members on the same machine communicate reliably through its operating system session UNDERLAY (lower e b layer) a d to set up this link/session: 1 A sends request to a 2 a looks up registration of E , finds e 3 a sends request to e 4 e sends request to E

  11. A NEW LAYER MODEL: THE MAJOR COMPONENTS PROTOCOLS STATE ALGORITHMS or, the can be centralized or, the “data plane” or distributed “control plane” across the members in any way where members members member algorithm are registered in underlays attachments attachment algorithm registrations of overlay members in locations location algorithm this layer session protocol sessions links link algorithm forwarding protocol routes routing algorithm

  12. A NEW LAYER MODEL: SCOPE AND LEVEL layers are arranged in a the scope of a layer is the set or class “uses” hierarchy, which of processes that could be members defines levels application process APPLICATION LAYERS IP interface this describes of machine the classic Internet architecture INTERNET 1 2 in terms of the CORE new layer (IP, TCP, UDP) model or “geomorphic view” LANs 1 2 Ethernet interface gateway

  13. WE CALL THIS THE “GEOMORPHIC VIEW” OF NETWORKING . . . . . . BECAUSE THE COMPLEX ARRANGEMENT OF LAYERS RESEMBLES THE EARTH’S CRUST

  14. HOW THE GEOMORPHIC VIEW IS DIFFERENT FROM SDN The geomorphic view The way that SDN has been described so modularizes the complexity of far, all the action is stuffed into a single, real networks, spreading it out large network layer. over multiple, relatively simple layers. Even if the implementation looks like this, the geomorphic view is a better abstraction for understanding the requirements that the implementation should satisfy.

  15. OUTLINE 1 THE “GEOMORPHIC VIEW”: “ AN ABSTRACT MODEL OF NETWORK ARCHITECTURE 2 THE GEOMORPHIC VIEW OF MOBILITY a Understanding mobility b Comparing existing mobility protocols c Generating and composing new mobility mechanisms 3 BEYOND MOBILITY: FUTURE WORK

  16. NETWORK MOBILITY IS . . . . . . THE HOT THING . . . A SOURCE OF CONFUSION it’s a mobile world many dozens of proposals devices move from one medium to another, as well as from place to place there is no sound basis for qualitative or quantitative ubiquitous computing (under the new evaluation of them name “machine-to-machine”) is an area of enormous growth they are not designed to be composed with one another . . . VERY DIFFICULT TO IMPLEMENT AT INTERNET SCALE despite the number of proposals, there are gaping the Internet has a hierarchical holes in their coverage address space (partly geographical) the Internet scales because subtrees are address blocks, and address blocks are aggregates for routing mobility breaks the rules—an individual IP address moves to a place where it does not belong

  17. MOBILITY AND LAYERS “mobility”: this is true, but not enough! a mobile device maintains network connectivity as layers explain it moves from place to place the identity, persistent credentials, identifier and capabilities that are retained while old new something location location else changes also, the strongest definitions of mobility include continuity of a channel (instance of a communication service)—most such channels are inter-layer higher endpoint BENEFITING LAYER service specification LAYER IMPLEMENTING lower endpoint MOBILITY

  18. THERE ARE TWO PATTERNS FOR IMPLEMENTING MOBILITY BENEFITING LAYER DYNAMIC- A B ROUTING MOBILITY this link connects LAYER a to the rest of its IMPLEMENTING layer MOBILITY a b1 a1

  19. THERE ARE TWO PATTERNS FOR IMPLEMENTING MOBILITY BENEFITING LAYER DYNAMIC- A B ROUTING MOBILITY LAYER as the attachment IMPLEMENTING of a member MOBILITY changes, its links change, a b1 and the routing algorithm must find new routes to it old link new going link away imple- mented layer state by components this that change: layer attachments links a2 routes

  20. THERE ARE TWO PATTERNS FOR IMPLEMENTING MOBILITY BENEFITING LAYER DYNAMIC- SESSION- A B ROUTING LOCATION MOBILITY MOBILITY LAYER as the attachment IMPLEMENTING of a member MOBILITY changes, as part of the its links change, session state, a b1 and the routing a knows b1 as algorithm must the far find new routes endpoint of to it the session layer state components that change: attachments links a2 routes

  21. THERE ARE TWO PATTERNS FOR IMPLEMENTING MOBILITY BENEFITING LAYER DYNAMIC- SESSION- A B ROUTING LOCATION MOBILITY MOBILITY as the channel LAYER as the attachment endpoint changes IMPLEMENTING of a member its location MOBILITY changes, in the its links change, implementing a b1 b2 and the routing layer, algorithm must the session find new routes state changes to it to match it tell a that session endpoint is now at b2 layer state layer state components components that change: that change: attachments locations links sessions a2 routes

Recommend


More recommend